Hierarchical Fusion Image Enhancement Algorithm Based On Retinex

In the field of digital image processing,Retinex algorithm as one of the most common image enhancement algorithms,based on the characteristics of human visual system.It has the characteristics of maintaining color constancy,the ability of not easily distorted,local contrast enhancement and dynamic change of image in an ideal range.And it has broad application prospects in all ranges of works and has become a research hotspot in this field.Aiming at the problem of insufficient edge detail enhancement in Retinex algorithm and unable to control the enhancemnt degree of information effectively at different scales,a multi-level decomposition algorithm for Retinex image enhancement is proposed.Firstly,in Retinex algorithm,combined with rotary symmetric bilateral filtering,reflective components and illumination components representing different scales of image are obtained by setting different filtering parameters.Secondly,all levels of reflective components are enhanced by using exponential function,and the final illumination components are processed by combining Sigmoid function and gamma correction,which can effectively enhance the illumination components.The expressive ability of image edge details improves the overall brightness of low-illumination image and suppresses the high-brightness region.Thirdly,the fusion of enhancement components at different levels combines the fused image brightness map with the original color component to make the color saturation consistency of the adjusted image better.Finally,the enhanced image is processed by color restoration function to further avoid color deviation and distortion.Compared with six classical image enhancement algorithms,the experimental results show that the average value of local standard deviation,information entropy and average gradient are improved to some extent,which verifies the robustness and effectiveness of the proposed algorithm.The paper has 32 figures,3 tables and 65 references.